This is the online version of Adam’s Biotech Scorecard, a subscriber-only newsletter. STAT+ subscribers can sign up here to get it delivered to their inbox. STAT is also producing a free pop-up newsletter about this year’s ASCO meeting. Sign up here.
Chicago, here I come. The annual meeting of the American Society of Clinical Oncology kicks off tomorrow and runs through Tuesday, June 2.
Let’s set the table for cancer research’s big event, starting with the obvious headliner: Revolution Medicines and its RAS-blocking pancreatic cancer drug daraxonrasib. Full results from the company’s Phase 3 RASolute 302 study will feature prominently at ASCO’s plenary session on Sunday afternoon. The arena-sized exhibition hall will be jam-packed.
Although gene editing has enormous clinical promise, it still faces many obstacles that must be overcome before broad translation. For example, the genome editing field is continuously working to increase the safety of the technique. One way to do that is to control the duration of gene editing activity. However, achieving precise temporal control over these platforms is challenging.
Scientists currently lack tools that can precisely control the duration of gene editing therapies, to halt their effects after a few months or years. Researchers have developed several potential solutions, including degradable protein- or RNA-based systems, but existing approaches face limitations such as only being applicable in the liver.
Now, a new gene editing system, named PRINCE, with inducible nuclease proteins and guide RNAs, enables researchers to control the duration and specificity of gene therapies precisely with small molecules—potentially addressing a longstanding safety concern in the field.
In the PRINCE CRISPR-Cas gene editing system, expression of the nuclease and the guide RNA is separately inducible by two approved small molecule drugs, enabling both temporal control and reduced off-target editing.
The platform was stable for as long as two years after genomic integration in cultured human cells. In addition, a smaller system, called Little Prince, showed promising signs of efficacy in humanized mouse models of elevated cholesterol levels and age-related macular degeneration. Little Prince uses compact nucleases that can be delivered in a single adeno-associated viral vector (AAV) for delivery.
In two mouse models, Little Prince reduced excessive cholesterol levels in mice with genetic hypercholesterolemia and showed signs of reducing lesion size in rodents with laser-induced choroidal neovascularization—a model of age-related macular degeneration.
More specifically, Little Prince “ameliorated pathological phenotypes of hypercholesterolemia (average reductions of 45% and 47% in serum total cholesterol and low-density lipoprotein cholesterol, respectively) and neovascular age-related macular degeneration, with significantly reduced lesion size and leakage (P < 0.0001).”
The system produced fewer off-target edits and lower off-target editing frequencies than constitutive nuclease expression, highlighting the utility of precise temporal control.
“PRINCE and Little Prince also provide […] capabilities that might also be useful for research objectives that include lineage tracing and conditional genetic engineering work,” the authors noted. These results, they asserted, position PRINCE and Little Prince as controlled genome editing platforms with potential for in vivo, particularly in situ, therapeutic applications.
From the revived corpse of Frankenstein’s monster to the disembodied hand, “Thing,” in the Addams Family, reanimated tissue is one of the most enduring images in science fiction. The discovery of a sea floor-dwelling sea cucumber that scientists are calling a “real-life zombie” suggests that there may be some basis for that image in nature.
Scientists headed by a team at Memorial University of Newfoundland showed the continued viability of amputated tissue from the sea cucumber Psolus fabricii for more than three years in natural seawater. It’s the first known report of the long-term survival—and continued growth—of discarded tissue outside of a highly controlled, sterilized environment.
The discovery that these living P. fabricii explants (LiPfe) can survive for years in natural seawater without any supplementation challenges assumptions of what’s possible for tissue immortality and could have implications in areas including regenerative biology and tissue engineering. The findings could also lead to the development of experimental models for biological research that are more widely accessible, without the ethical and logistical challenges associated with many existing cell lines.
“We haven’t grown a new, complete sea cucumber yet, but we are seeing pretty stunning growth and diversification of cells literally years after this tissue was removed,” said research lead Rachel Sipler, PhD, a Bigelow Laboratory for Ocean Sciences senior research scientist. “It’s like a lizard that loses its tail. We know some lizards can grow new tails; we’re talking about whether the tail can grow a new lizard.”
Reporting on their findings in Science Advances (“Natural tissue immortality: Indefinite survival of sea cucumber explants,”) Sipler and colleagues stated, “Our findings challenge conventional perceptions of tissue immortality and present a new class of experimental model, free from ethical concerns, with substantial implications for regenerative biology, biomedical research, and tissue engineering.”
Over the last 200 years, scientists have tried to achieve cellular and tissular survival outside living hosts, “… but efforts have been met with limited success due to the highly degradable nature of tissue itself,” the authors wrote. Since the mid-20th century, scientists have made significant breakthroughs with immortal cell lines, such as HeLa cells, that can be grown in a lab and proliferate indefinitely for long-term research. In earlier studies, tissue cultures have only been maintained under axenic conditions that are tightly controlled, rigorously maintained, and lack any bacteria or other organisms. Even then, they have not demonstrated signs of actual healing and growth, nor retained the ability to move independently. “While immortal cell lines demonstrate indefinite proliferation in vitro, they lack structural integrity and complex tissue interactions,” the team continued. “Achieving this with complex, structured tissue represents the next step.”
Many echinoderms, including sea cucumbers, are known to display impressive regeneration capacity and negligible cell aging. “In the ongoing effort to understand tissue culture, regeneration, and immortality, researchers have naturally been drawn to echinoderms, a phylum with genetic and evolutionary links to vertebrates and examples of both extreme regenerative capacity and negligible cellular senescence,” the investigators noted. Lost tissue, though, was always assumed to eventually decay or die.
Yet, in what Sipler calls a product of “keen observation,” the researchers noticed that some discarded tissue from a tube foot of a sea cucumber hadn’t decayed after a number of weeks. In fact, it seemed to be growing. The researchers then ran a number of experiments in flowing seawater with tissue removed from the feet, main body, and tentacles of three individuals of P. fabricii, a cold-water species of sea cucumber.
They found evidence of diversifying cells, immune activity, and tissue reorganization in the explanted tissue. “In experimental trials, these explants, termed LiPfe (living immortal P. fabricii explants), displayed immune activity, cell cycling, tissue reorganization, and absorption of dissolved amino acids, underscoring their active living state,” they noted. And in the absence of a mouth, the cells appeared to be getting nutrients by absorbing amino acids dissolved in the seawater.
Even after three years, when the researchers stopped the experiments in order to publish, the tissue was still active. This ability to survive in a complex, stressful environment, Sipler said, makes this cell line unique compared to other tissue cultures. “Compared to other cells or tissues grown under laboratory setups that required strict parameters, including axenic conditions, LiPfe required nothing apart from natural running seawater,” they wrote. “Comparative experiments conducted on explanted tissues from related species demonstrated no equivalent tissue survival, highlighting the unique properties of P. fabricii, which do not have parallels in the current literature.”
“Natural seawater is just about the most microbially diverse, least clean approach we could take experimentally,” Sipler added. “Yet, that rich environment full of bacteria and all this organic matter was actually feeding them and allowing this tissue to heal and grow.”
The implications for biomedical sciences and engineering, the authors said, are profound, with potential applications in everything from tissue regrowth to anti-microbial healing. In their paper, the authors stated, “The discovery of LiPfe challenges the boundary between organismal life and cellular autonomy, compelling a redefinition of what it means for tissue to be alive.”
The discovery opens up new opportunities for biological research and education more broadly. The tissue they’ve preserved not only shows an unprecedented ability to maintain its structural integrity and complexity in culture. It can also be grown more easily in the lab and, as an invertebrate, isn’t subject to as many research restrictions, making it useful in contexts where there are legal obstacles or limited biosafety infrastructure for using human-based or other vertebrate cell lines.
As an oceanographer, Sipler noted that the exciting discovery drives home the incredible untapped potential of ocean life. “The best advances in science are made when you find a natural analog for what you’re studying,” she said. “Here is this species that has this groundbreaking ability, and we had no idea. It’s a reminder of how much is yet to be discovered in the marine environment, and how important it is to protect these resources that may hold really valuable knowledge for us.”
The remarkable clinical success of CAR T therapies in blood cancers has validated the promise of engineered cell therapies. But according to Adam Janvier, PhD, head of cell therapy at eXmoor Pharma, the industry faces a crucial challenge: transforming highly personalized, labor-intensive manufacturing into scalable, commercially viable production systems.
“The next big thing that really needs to come through is how we can start tackling solid tumors,” Janvier says. “Until solid tumor is truly tackled, we’re always going to be missing that key next step of what CAR T has the promise to do.”
Although scientific hurdles remain, Janvier emphasizes that manufacturing constraints are equally pressing. Today’s autologous CAR T therapies rely on harvesting a patient’s own immune cells, engineering them outside the body, and reinfusing them after a manufacturing process that can stretch to two weeks. For critically ill patients, that timeline can prove devastating.
“We might fail a process because the donation just isn’t good enough,” he explains, referring to inconsistent starting material collected from heavily pretreated cancer patients. “Then we’ve got a four to six week vein-to-vein time due to manufacturing, testing, and logistics, where the patient might pass away during the period.”
The dual risks of manufacturing failure and lengthy turnaround times are pushing developers and contract development and manufacturing organizations (CDMOs) toward alternative strategies. Among the most promising are allogeneic, or off-the-shelf, CAR T therapies and emerging in vivo approaches that could eliminate ex vivo manufacturing altogether.
“There’s a lot of work going on now with in vivo CAR T,” Janvier says. “Instead of taking a blood donation as starting material, there is growing evidence that we could use the reprogramming technology directly with the patient, generating functional CAR T cells in situ.” Although such approaches remain early-stage, they represent a potential paradigm shift by reducing manufacturing time, simplifying logistics, and lowering costs.
Analytics and quality control also remain major bottlenecks. Current CAR T testing workflows rely heavily on expensive, time-consuming assays, including flow cytometry, qPCR, and tests to confirm the quality and safety of the lentivirus. Janvier believes that AI could eventually streamline many of these processes.
“One of the exciting technologies coming out is AI-based flow-cytometry approaches,” he says, pointing to emerging platforms that use label-free imaging and machine learning to characterize cells without fluorescent antibodies. “All of a sudden, you’re removing the need for antibodies and fluorophores, reducing the cost,” he says.
Still, Janvier argues that automation might ultimately become the defining factor in whether CAR T therapies can achieve widespread commercial adoption. Current cleanroom manufacturing remains highly manual, requiring specialized staff and flexible—but inefficient—facility layouts. “Once we approach commercial scale, batch costs need to have substantially decreased,” he says. “Automation can really support that.”
Janvier envisions future CDMOs operating sophisticated robotic manufacturing platforms capable of running around the clock while minimizing operator variability and contamination risk. However, implementing such systems will require substantial capital investment and new technical expertise. “These are not going to be inexpensive methods to implement into facilities,” he notes. “They’re going to be a large capital investment, and also a large people investment.”
Beyond manufacturing hardware, Janvier believes structural changes must begin much earlier in therapy development. Many CAR T programs originate in academic laboratories focused primarily on biological innovation rather than manufacturability or commercial scalability. “What can be missed there is the translation starting at the very beginning,” he says. “You need to start with the end in mind.”
That means considering GMP compatibility, scalability, cost-of-goods analysis, and automation readiness long before therapies enter clinical trials. Investors, Janvier adds, are increasingly demanding evidence that therapies can ultimately be manufactured at scale—not simply that the science is compelling.
A maturing, data-rich and more commercially-focused cell and gene therapy (CGT) industry is finally embracing innovative manufacturing technologies, according to the head of U.K.-based industrial software firm, Autolomous.
Early CGT manufacturing relied heavily on technologies borrowed from biologics manufacturing, systems originally built for producing therapeutic proteins rather than patient-specific therapies. As a result, production processes were often highly manual, fragmented, and operator-dependent. Indeed, in many cases, manufacturing resembled a specialized laboratory process, rather than a scalable industrial operation.
But the situation is changing, says Autolomous CEO, Alexander Seyf, who argues CGT manufacturers are embracing innovation with product quality, cost control, and scalability in mind.
“As the industry moves beyond early clinical programs and toward commercialization, manufacturers are recognizing that traditional approaches simply won’t support the scale, consistency, or economics needed long term.
“There’s now much greater focus on automation, closed processing, and digital integration to improve reproducibility, reduce contamination risk, and lower the cost of goods,” Seyf tells GEN.
Increasingly, CGT firms are opting for purpose-built technologies, he says, citing automated cell handling, closed-system processing, robotic fill-finish, and integrated single-use platforms as examples.
“The broader shift is toward what many describe as ’CGT 4.0:’ applying the principles of Industry 4.0 to cell therapy manufacturing. The aim is to move away from highly manual, artisanal production toward scalable, repeatable manufacturing that can support broader patient access,” he adds.
Data
Data is also driving the adoption of new technologies. In cell and gene therapy production, large volumes of data are collected by different instruments, software platforms, QC systems, environmental monitoring tools, and manual inputs—many of which were never designed to communicate with one another.
In such circumstances, digital technologies provide manufacturers with an infrastructure that can gather process information and ensure it is usable, traceable, and reliable, Seyf says.
“Maintaining data integrity and full chain-of-custody visibility across fragmented systems can become incredibly difficult, particularly as operations scale.
“To solve this, companies are investing heavily in integrated digital architectures that bring manufacturing and quality data together into unified environments. Standardized data models, interoperable software platforms, and automated data capture are becoming increasingly important. Cloud infrastructure also plays a key role because it allows manufacturers to aggregate and analyze data across multiple facilities in real time,” Seyf says.
A typical, modern digital cell and gene therapy manufacturing setup includes systems that manage manufacturing execution, laboratory information, quality, electronic batch records, and cloud-based data platforms.
“Together, these systems help manage scheduling, batch tracking, compliance, traceability, and quality oversight in real time,” Seyf says, adding, “Modern facilities are also increasingly using process analytical technologies and integrated sensors to monitor critical process parameters continuously, rather than relying only on end-point testing.”
“A digital CGT manufacturing system is really about connectivity and visibility across the entire process. It’s not just about replacing paper records with electronic systems—it’s about creating an environment where manufacturing equipment, quality systems, analytics, and logistics are all connected and continuously sharing data.”
Material traceability
The need to keep track of cell and gene therapy raw materials is also fueling the adoption of innovative technologies, with patient-specific therapies being a case in point.
Seyf tells GEN, “Maintaining chain-of-identity and chain-of-custody is particularly important in CGT manufacturing, especially for autologous therapies where every batch is tied to an individual patient. That requires seamless integration between manufacturing systems, analytics platforms, and logistics operations.”
A wide range of new technologies offers the next generation of oligonucleotide manufacturers a choice in how they will commercially manufacture their products. That’s the view of a technical expert with experience working with small biotechs and large pharmaceutical companies.
According to Barrie Cassey, PhD, a former technical director at GSK, the tide of oligonucleotide therapies now reaching the clinical stage is driving companies to commit to cheaper high-production techniques, such as enzymatic ligation.
“What’s driving this technology is the ability to make [oligonucleotide therapies] at scale,” he says.
According to Cassey, most oligonucleotide drugs are currently produced by solid-phase techniques, but the capacity for commercial manufacturing is limited globally, with most therapies being for rare disease patients. However, looking at products in Phase III, 40% target large patient populations, Cassey says, and this rises to around 50% for products in Phase II.
“If you extrapolate, assuming 70% of these Phase III assets become commercial (which is pretty conservative), your total demand could rise to six to seven tons globally, which is more than all global capacity today,” he explains.
Overall, within eight years, global manufacturing demand could reach as high as 30 tons—far outstripping current manufacturing capabilities, Cassey says. Moreover, drugs targeting large patient populations (e.g., statins) have traditionally needed to be cheap to manufacture, he says.
To get around this, large pharmaceutical companies are increasingly looking at new techniques for oligonucleotide manufacture, such as experimenting with fragmentation versus whole molecule approaches. One company has invested $250 million in a manufacturing plant using enzymatic ligation, Cassey adds.
Overall, Cassey says, the take-home message is that new techniques are closer than many people think, with some companies already offering them for preclinical manufacture.
“These technologies are almost ready to go and, if someone wanted to use them in twelve months, it might be possible,” he says. “It’s just about getting the right conditions for everything to coalesce together.”
A broad European industry/academic consortium, Inno4Vac, has developed a robust, open-source modeling platform that unifies upstream and downstream models and predicts product stability, all within a single environment. Unlike many open-source modeling programs, this one, called CADET-Hub, doesn’t need mad programming skills to use.
As Reinhard Liebers, PhD, a project manager at the European Vaccine Initiative, tells GEN, this work shows “that mechanistic modeling, digital twins, and stability prediction can be integrated into one collaborative platform to accelerate bioprocess development, improve process understanding, and reduce experimental workload.”
CADET-Hub is an integrated, modular, cloud-based modeling platform designed to help the biotech industry especially vaccine manufacturers quickly reconfigure their manufacturing platforms in response to emerging demands such as quickly-mutating pathogens. Its developers add that it can be used to help optimize a broad spectrum of bioprocesses.
Integration and ease of use—even for non-programmers—are standout features for this new platform. A key motivation for developing CADET-Hub was “to make the setting up and fine-tuning of biomanufacturing processes more efficient as well as reduce the technical overhead of setting up local software environments for bioprocess modeling,” according to a team led by Eric von Lieres, PhD, head of modeling and simulation at Forschungszentrum Jülich, and Liebers, in a recent paper. Therefore, each of the modules is web-based “in principle, with options for local deployment if desired, and a graphical user interface under development,” they tell GEN.
The upstream models are based on computational fluid dynamics and metabolic models for bioreactor simulations. The downstream models offer insights into centrifugation, filtration, and chromatography. CADET-Hub includes a dedicated stability forecasting model.
The research around CADET-Hub supports a shift from empirical trial-and-error development toward predictive, data-driven biomanufacturing with integrated upstream, downstream, control, and stability models,” Liebers says.
Overview of CADET-Hub and the broader CADET-Hub environment, which integrates standalone modeling, the open-source CADET framework within a JupyterHub-based online platform. [Inno4Vac]
Case studies
The team of authors around Liebers and von Lieres tested CADET-Hub under three scenarios.
As a way to model anion exchange capture of a recombinant subunit vaccine, they report that it functioned effectively as a collaborative workflow to share experimental data and mechanistic models, capturing meaningful nonlinear and multicomponent effects.
When modeling downstream filtration using ion exchange chromatography, CADET-Hub used explicit storage vessel models to close local material balances and transfer component concentrations in a structured way. It showed how upstream variability “can be propagated downstream to predict its impact on subsequent unit operations.” But, they add, “Experimental validation of the filtration model is still required.”
For process optimization and control, they point out, “Vaccine production requires a combination of sequence control and nonlinear model predictive control.” Overall, the case study showed a 50% improvement in upstream yield and a 20% improvement in downstream productivity that is attainable through model-based optimization.
To predict product stability under various conditions, the scientists combined hierarchical and advanced kinetic models within a Bayesian framework. They produced a “robust estimation of vaccine shelf life,” while meeting regulatory expectations and ICH Q12 principles.
“The next steps are to expand CADET-Hub toward fully end-to-end digital bioprocess models; to improve scalability, industrial deployment, and user interfaces; and to continue regulatory validation and adoption,” Liebers says.
CADET-Hub was created as part of the Inno4Vac vaccine development project and received funding from the Innovative Medicines Initiative/European Union/European Federation of Pharmaceutical Industries and Associations (IMI2/EU/EFPIA).
Want to stay on top of the science and politics driving biotech today? Sign up to get our biotech newsletter in your inbox.
Good morning. Today, we have some AI drug development news and an examination of the growing longevity industry.
The need-to-know this morning
The FDA has pushed back the decision date for AstraZeneca’s experimental breast cancer drug camizestrant, following a negative vote from a group of agency advisers, the company said. The extra time will allow the FDA to review additional analyses that AstraZeneca is providing. FDA advisers took issue with the study design of the pivotal SERENA-6 trial, though European regulators have recommended the drug be approved. AstraZeneca did not specify the new target date for an FDA decision.
Blackstone Life Sciences, a private equity fund, is providing Apogee Therapeutics with up to $1.3 billion to pay for the Phase 3 development and potential commercialization of zumilokibart, the biotech’s long-acting treatment for atopic dermatitis.
Kailera’s own ‘triple-G’ drug also looks very powerful
Kailera said yesterday that its investigational obesity drug that targets three hormones led to significant weight loss in a Phase 1 study.
Biohub, the non-profit research organization co-founded by Priscilla Chan, MD, and Mark Zuckerberg, has now unveiled the latest update to the ESM protein language model family, with expanded capabilities in binder design and protein function mapping for therapeutic discovery.The release comes just seven months after Biohub recruited the team behind EvolutionaryScale.
The system includes ESMC (Evolutionary Scale Modeling Cambrian), a language model trained on approximately 2.8 billion sequences drawn from a breadth of life, including organisms adapted to extreme environments, and more than 20,000 types of proteins found in the human body. Evolutionary information encoded in ESMC is translated into atomic-resolution protein structures and interactions using the design engine and prediction model, ESMFold2.
Alex Rives, PhD, head of science at Biohub and former chief scientist at EvolutionaryScale, presented the work at this week’s “AI in Biology” symposium at Cold Spring Harbor Laboratory.
These models aim to transform the earliest stages of drug discovery by making biology more programmable. While traditional discovery workflows rely on slow and resource intensive experimental screens to identify promising drug candidates, rational protein design guided by in silico predictions has the potential to dramatically accelerate development timelines.
“We’re at an exciting point in protein biology where accurate digital representations allow asking experimental questions at a scale that wouldn’t be possible in the laboratory,” Rives told GEN Edge.
ESMC provides a foundation for modeling the sequence, structure, and function of proteins. ESMFold2 predicts the structure of proteins and biomolecular complexes. Features derived from the representations of the model capture fundamental principles of structure and function that form a compositional grammar for protein biology. [Biohub]
ESMFold2 designed high-affinity protein binders against five disease targets in cancer and immunology: receptor tyrosine kinases implicated in tumor growth (EGFR and PDGFRβ), immune checkpoints exploited by cancer cells to evade immune surveillance (PD-L1 and CTLA-4), and a regulator of immune cell signaling (CD45).
Lab-validated designs achieved hit rates ranging from 36–88% for compact mini-binders and 15–29% for antibody-derived formats, while also demonstrating nanomolar binding affinity, high specificity, and favorable stability profiles consistent with potential clinical utility. Notably, binders for PD-L1 showed therapeutic function and restored T-cell signaling in laboratory tests by blocking the same pathway as approved checkpoint therapies.
Rather than requiring multiple sequence alignments (MSAs) to build representations, ESMFold2 captures evolutionary information encoded during pretraining. The model also uses a looped transformer architecture, which allows compute to scale at inference time and avoids overfitting that can arise when training is constrained by limited experimental protein structures.
In benchmarking, ESMFold2 performed favorably when compared against Chai-1 from Chai Discovery, Boltz-1 from MIT (whose developers have since launched a public benefit corporation), and AlphaFold 3 from Google DeepMind.
The models are accessible under the highly permissive Massachusetts Institute of Technology (MIT) license for both commercial and non-commercial use. The work is described as a preprint that has not yet been peer reviewed.
“All in” on AI biology
Last November, Chan and Zuckerberg made the pledge to go “all in on AI-powered biology,” announcing that the organization’s scientific teams will now unite under a single entity, known as Biohub, where the duo would place the majority of their philanthropic effort.
Concurrently, Rives and EvolutionaryScale colleagues were recruited to tackle disease by decoding the “grammar” of amino acids through billions of years of evolution. That same mission had once secured a whopping $142 million seed round when the startup unveiled in 2024. The raise was led by Nat Friedman, Daniel Gross and Lux Capital, and included participation from Amazon Web Services (AWS) and NVentures, Nvidia’s corporate venture arm.
Biohub continues to advance virtual biology by building digital representations of molecules, genomes, cells, and living systems. The new ESM release joins Biohub’s growing ecosystem of biology models, including TranscriptFormer, which was published in Science earlier this month.
The organization recently invested $500 million in the Virtual Biology Initiative, a five-year campaign to accelerate the creation of technologies and multi-modal datasets to build predictive models of biology. The commitment comes a few months after the organization announced a collaboration with Arc Institute and Tahoe Therapeutics to build the largest single cell chemical perturbation dataset to power the virtual cell.
Evolution is all you need
Biohub has applied the new ESM models to generate theESM Atlas, a mapping of 6.8 billion sequences and 1.1 billion predicted structures to protein function using ESMC’s representations. Generating this atlas would have taken “billions of years of experimental work,” but was condensed into a couple of weeks with computational inference. The ESM Atlas is released open-source.
By probing ESMC’s representations using sparse autoencoders (SAEs), a technique for identifying interpretable structure in large language models, the authors found that the model independently learned hierarchical organization covering the basic chemistry of individual amino acids, local structural interactions, and functional concepts across unrelated proteins, despite being trained only on sequence data.
Notably, ESM Atlas SAE feature clusters brought together RNA-guided DNA endonucleases, eukaryotic Fanzor proteins and their evolutionary ancestor, prokaryotic TnpB, despite their high evolutionary divergence and low sequence similarity. These insights could support the development of new gene-editing tools.
While the preprint’s results are still a step away from clinical impact, Rives reiterates the power of open science in placing these tools in the hands of researchers working directly in translational research.
Biohub is partnering with a number of platform partners, including AWS Bio Discovery, Benchling, Phylo, Tamarind Bio, Modal, Tool Universe, and SandboxAQ, to make the models widely available.
![ESMC provides a foundation for modeling the sequence, structure, and function of proteins. ESMFold2 predicts the structure of proteins and biomolecular complexes. Features derived from the representations of the model capture fundamental principles of structure and function that form a compositional grammar for protein biology. [Biohub]](https://www.genengnews.com/wp-content/uploads/2026/05/Viz1-11-1024x576.png)